Second Function (warfare). One cannot retrieve a single name of a Proto-
Indo-European war god. A proposed lexical correspondence (that would yield
a PIE *ma
¯
wort-) between the names of the Latin war god Ma
¯
rs and the Skt
Maruta
´
s is doubtful; the latter are companions of the war god Indra. Rather we
have, with the exception of Indo-Iranian, a series of diVerently named war
gods: Skt Indra $ Skt (Maha
¯
bha
¯
rata) Arjuna $ Av Indara $ Lat Ma
¯
rs $ Lat
(Livy) Tullus $ ON Tho
¯
rr $ Gaul Taranis $ OIr Ogma.
The second function can also be viewed in terms of two aspectually contrast-
ing warrior functions—: defensive (good) and oVensive (wild, destructive to the
community itself)—and this opposition is seen to be played out among some of
the pantheons. The more destructive manifestations are seen in the following
correspondences: Skt (Vedic) Va
¯
yu (a storm god) $ Skt (Maha
¯
bha

¯
rata) Bhı
¯
ma
$ Av Vayu.
Thunder god (*perk
w
unos). The lexical set consists of ON Fjo˛rgyn, Lith
Perku
¯
´
nas, ORus Peru
´
nu
00
, and perhaps Skt Parja
´
nya. The underlying root is
probably *per- ‘strike’ with diVerent extensions built in diVerent groups. The
North-West European set is relatively coherent with associations with the
thunder god (Fjo˛rgyn was the mother of the Norse thunder god Tho
¯
rr), hurling
lightning, use of the club both in battle but also as a fertility symbol at
weddings. The association of the North-Western deities with the Sanskrit
deity is not so clear, although the latter is depicted as a rain god in the Vedas.
Third Function. No lexical correspondence here but rather a series of gods
who Wnd themselves third in canonical order of deities and who are associated
with fertility. These may especially include the divine twins but also single
deities such as Lat Quirinus or ON Freyr, Gaul Teutates and OIr Bres.

Transfunctional goddess. There is no lexical evidence for such a deity but the
diVerent Indo-European traditions are replete with examples of goddesses
whose qualities either comprise or dispense the three functional categories.
Such goddesses may be provided with a trifunctional epithet, e.g. the name of
the Iranian goddess Ar@dvi Su¯ra Ana
¯
hita
¯
may be rendered ‘moist, strong, and
pure’ just as Athena is showered with the epithets po
´
lias, nı
´
ke
¯
, and hugı
´
ea
‘protectress, victory, well-being’ and Juno is Seispes Ma
¯
ter Regı
¯
na ‘safe,
mother, queen’, in all cases—although not necessarily in canonical order—
words suggesting the three Dume
´
zilian functions. We have already seen how
the three functions may also be split among three associated goddesses, e.g. the
Greek judgement of Paris where Hera promises rulership, Athena military
victory and Aphrodite oVers the love of the most beautiful woman, or the

three semi-divine Machas of early Irish literature.
Aryan god (*h
4
ero
´
s). A deity in charge of welfare is indicated by a number
of lexical correspondences (Skt Aryaman,Avairyaman, Gaul Ariomanus, OIr
Eremon, and non-cognate functional correspondences, e.g. Vidura in the
25. COMPARATIVE MYTHOLOGY 433
Maha
¯
bha
¯
rata. The Aryaman-type deity is associated with the building and
maintenance of roads or pathways, with healing, especially involving a ritual
where cattle urine or milk is poured in a furrow, and the institution of
marriage. In this sense he is seen as a ‘helper’ to the First Function deity of
the Mitra type.
In addition to these there are a number of deities that have been proposed
either on the basis of limited isoglosses (Greek-Sanskrit) or on questionable
linguistic evidence.
Pastoral god (*pe
´
h
2
uso
¯
n). Primarily a Greek (Pa
¯
´

n)-Sanskrit (Pu
¯
s
_
a
¯
´
) corres-
pondence, possibly from *peh
2
- ‘protect, feed (cattle)’. Both deities are pas-
toral gods and are closely associated with goats. In Greek mythology some of
Pan’s original characteristics may also have been assimilated by his father
Hermes.
Medical god. Both the Indic god Rudra and Greek Apollo inXict disease
from afar by their bows and are also known as healers; both are also associated
speciWcally with rodents, Rudra’s animal being the ‘(rat) mole’ and Apollo was
also known as Smintheus ‘rat god’.
Decay goddess. This is based on an Indic-Latin isogloss where both tradi-
tions indicate a goddess (Skt Nı
´
rr8ti-, Lat Lu¯a Mater) whose names derive from
verbal roots ‘decay, rot’ and are associated with the decomposition of the
human body.
Wild god (*rudlos). The only certain deity by this name is the Skt Rudra
´
-
although there is an ORus Ru
˘
glu

˘
(name of a deity) that might be cognate.
Problematic is whether the name derives from *reud- ‘rend, tear apart’ as Lat
rullus ‘rustic’ or from the root for ‘howl’.
River goddess (*deh
a
nu-). This is largely a lexical correspondence, e.g. Skt
Da
¯
nu, whose son holds back the heavenly waters, and Irish Danu, Wels Do
ˆ
n,
both ancestor Wgures. The same root underlies the names of many of Eur-
ope’s larger rivers, including the Danube, Don, Dnieper, and Dniester (the
latter three as Iranian loans). Other than the deiWcation of the concept of
‘river’ in Indic tradition, there is really no evidence for a speciWc river
goddess.
Sea god (*trih
a
to
¯
n). Even more doubtful is the Celtic-Greek possible cor-
respondence between OIr trı
¨
ath ‘sea’ and the Greek sea god Trı
¯
´
to
¯
n, the son of

Poseido
¯
n. The lexical correspondence is only just possible and with no
evidence of a cognate sea god in Irish (there are other sea deities but these
are not lexically cognate), there is really no certain evidence of a god of
the sea.
Smith god (*wl8ka
¯
nos/*wl8keh
a
nos). This is based on a linguistically doubtful
comparison of the name of the Latin smith god Volca
¯
nus and the Ossetic smith
god wærgon. The problem here lies in the etymology of the Latin name which
434 25. COMPARATIVE MYTHOLOGY
may be derived from Etruscan or an Aegean loanword. There are no mytho-
logical elements, other than those generic to most smith gods, that might unite
the Latin and Iranian deities.
25.3 Creation
Although the various Indo-European groups exhibit diVerent creation myths,
there appear to be elements of a Proto-Indo-European creation myth pre-
served either explicitly or as much altered resonances in the traditions of the
Celts, Germans, Slavs, Iranians, and Indo-Aryans. These traditions all indi-
cate a proto-myth whereby the universe is created from a primeval giant—
either a cow such as the Norse Ymir or a ‘man’ such as the Vedic Purus
_
a—
who is sacriWced and dismembered, the various parts of his anatomy serving
to provide a diVerent element of nature. The usual associations are that his

Xesh becomes the earth, his hair grass, his bone yields stone, his blood water,
his eyes the sun, his mind the moon, his brain the clouds, his breath the wind,
and his head becomes the heavens. This body not only Wlls out the material
world but the dismemberment also provides the social tiers with the head
associated with the First (ruling) Function, the arms being equivalent with
the warrior function, and the lower torso, with its sexual organs, the fertility
function.
As to the identity of the sacriWcer we have hints in a related sacriWce that
serves as the foundation myth for the Indo-Iranians, Germans, and Romans
(with a possible resonance in Celtic). Here we Wnd two beings, twins, one
known as ‘Man’ (with a lexical cognate between Germanic Mannus and Skt
Manu) and his ‘Twin’ (Germanic Twisto, Skt Yama with a possible Latin
cognate if Remus, the brother of Romulus, is derived from *Yemonos ‘twin’).
In this myth ‘Man’, the ancestor of humankind, sacriWces his ‘Twin’. The two
myths, creation and foundation of a people, Wnd a lexical overlap in the
Norse myth where the giant Ymir is cognate with Skt Yama and also means
‘Twin’.
The dismemberment of the primeval giant of the creation myth can be
reversed to explain the origins of humans and we Wnd various traditions that
derive the various aspects of the human anatomy from the results of the
original dismemberment, e.g. grass becomes hair, wind becomes breath.
The creation myth is then essentially a sacriWce that brought about the
diVerent elements of the world. Conversely, as Bruce Lincoln has suggested,
the act of sacriWce itself is a re-enactment of the original creation. There is
evidence in various Indo-European traditions, e.g. Rome, India, that the parts
25. COMPARATIVE MYTHOLOGY 435
of the sacriWced animal were dispersed according to the prevailing social
patterns and, therefore, we may view the act of sacriWce as an attempt to restore
the balance of the world. This same notion may be carried also into the burial
ritual of at least some of the Indo-European traditions where it was imagined

that the deceased disintegrated back into its constituent parts, e.g. in the
R8gveda, the eye of the deceased goes back to the sun, his breath to the wind.
In a sense then, after the initial creation, life is essentially recycled.
25.4 War of the Foundation
This myth is attested primarily on the basis of Germanic (Norse) and Roman
sources but elements of it have also been claimed for Greek and Sanskrit. The
myth depicts the forceful incorporation of Dume
´
zil’s Third (fertility) Func-
tion into a social world run by the Wrst two functions. In Norse mythology,
the myth is expressed as a war between the Æsir, the gods of the Wrst two
functions, led by Oðinn and Tho
¯
rr, against the Vanir who were led by the
fertility gods Freyr, his sister Freya, and Njo
¨
rðr. After a period of warfare the
two sides conclude a pact of peace with the three fertility deities coming to
live among the Æsir, thus providing representatives of all three functions
within a single social group. The Roman parallel is found in the legend of
Romulus who, Wnding Rome lacking in women (fecundity), wars with the
Sabines. The Sabine women intercede and bring about peace between the two
sides and, again, the incorporation of the Third Function into society. The
Trojan War has also been interpreted in such light (the Greeks as the Wrst two
functions and the Trojans with Helen as the third). In Indic mythology, the
As
´
vins, representatives of the Third Function, Wnd their way into the world
of the other gods blocked by Indra until he is tricked into letting them in,
thus securing a three-function society.

25.5 Hero and Serpent
One of the central myths of the Indo-Europeans involves the slaying of a
serpent, often three-headed, by the archetypal hero, either deity or human.
Calvert Watkins has argued that this deed has left some lexical evidence in
the frozen expression *(h
1
e)g
w
he
´
nt h
1
o
´
g
w
him ‘he killed the serpent’, preserved
as such in Indo-Iranian with lexical substitutions in Hittite, Greek, and
Germanic. The association with three heads or some aspect of triplicity is
indicated either by descriptions of the monster, e.g. the three-headed dog
436 25. COMPARATIVE MYTHOLOGY
Ke
´
rberos who guards the Greek Underworld, the name of the hero, e.g. the
Skt Trita A
¯
ptya, or in some other aspect of triplicity, e.g. Horatio Cocles’
defeat of three opponents in early Roman history. Bruce Lincoln has sug-
gested that the context of this slaying is during the W rst cattle-raid where a
monster runs oV with the cattle of a hero whom he designates *Tritos ‘the

third’ who then sets oV in pursuit, accompanied by *H
a
ne
¯
´
r ‘Man’, kills the
serpent, and recovers his cattle. Traces of this myth are seen in Indo-Iranian,
Hittite, Greek, and Norse traditions.
25.6 Horse Sacrifice
It is largely the residue of ritual rather than explicit myths that points to the
existence of a speciWc association between the assumption of kingship and the
ritual mating with and sacriWce of a horse. The Indic as
´
vamedha, an inaug-
uration ceremony, and the Roman Equus October both involve the sacriWce of
a horse either to a warrior deity or on behalf of the warrior class; the victim
was a stallion that excelled on the right side of the chariot, and the victim was
dismembered, diVerent parts of the anatomy going to either diVerent loca-
tions or functionally diVerent deities. The medieval inauguration of an Irish
king in County Donegal which involved the king-designate bathing in
a cauldron with the dismembered pieces of a horse may also be a reXex.
The underlying myth, particularly in Indic, suggests some form of mating
between the king and the horse (mare), the latter of which behaves as a
transfunctional goddess and passes to the king the gifts of the three functions
that make up the totality of society.
25.7 King and Virgin
A recurrent theme, though not without considerable modiWcations (if genet-
ically inherited) or diVerences, is that of a virgin rescuing a king which is
found in Indic, Roman, Scandinavian, and Celtic sources. The basic structure
involves a king whose future (including his descendants) is endangered be-

cause of his immediate male relatives (sons, uncle, etc.) but is allowed to
prevail because of a virgin (often his daughter) who provides the oVspring
necessary to the king’s survival. In the Indic tale, for example, King Yaya
¯
ti is
rescued by four sons born to his daughter (who mated with three kings and
a teacher); in Roman tradition King Numitor’s line is ensured by the birth of
Romulus and Remus because his virgin daughter, Rhea Silvia, was made
pregnant by Ma
¯
rs.
25. COMPARATIVE MYTHOLOGY 437
25.8 Fire in Water
This mythic element is postulated on the basis of several disputed divine names
and some general mythic elements found in several Indo-European traditions
(Celtic, Italic, and Iranian). The lexical argument (Section 23.1) posits a PIE
*neptonos or *h
2
epo
¯
m nepo
¯
ts ‘grandson/nephew of waters’ on the basis of Skt
Apa
¯
´
mNa
´
pa
¯

t,AvApa˛m Napa
¯
t, and much less securely OIr Nechtain and Lat
Neptu
¯
nus. The myth itself depicts a divine being associated with Wre who
inhabits water (in the Celtic myth there is a sacred well of Nechtain whose
Wre burns out the eyes of those who approach it, in the Avesta the Wery power is
the xvar@nah, the burning essence of kingship, which was placed in Lake
Vourusaka) and who can only be approached by someone especially designated
for the task. Although there is no corresponding mythic evidence from Ger-
manic, the ON kenning sœvar niðr ‘son of the sea’, i.e. ‘Wre’, may provide some
linguistic support for the equation.
25.9 Functional Patterns
There are a number of patterns in Indo-European narratives that replicate the
three functions. Among the more striking are the motifs known as the ‘the sins
of the warrior’ and the ‘threefold death’. The Wrst motif deals with a represen-
tative of the Second Function whose downfall involves sins against all three
functions, e.g. the Germanic Starkaðr slays a king (violation of the First
Function), Xees in battle as a coward (violating his Second Function as a
warrior), and kills for money (a violation here taken to be against the third
estate). Traces of this motif also occur in other Indo-European traditions, e.g.
Greek where He
¯
rakle
¯
s manages three comparable sins or the Maha
¯
bha
¯

rata
where S
´
is
´
upa
¯
la commits three similar sins.
The ‘threefold death’ associates a particular type of death with a particular
function or functional deity. For example, classical sources indicate that among
the Gauls victims dedicated to the First Function Wgure (Esus) were hanged;
the Second Function (Taranis) received victims who had been burnt; and
victims dedicated to the Third Function (Teutates) were drowned. The motif
is also found in Germanic where the First Function deity, Oðinn, is known as
the ‘hanged god’ while victims to the fertility (Third Function) deity Nerthus
were drowned. These patterns are replicated in the heroic literatures of the
Celtic and Germanic peoples although the motif is believed to have been more
widespread. Essentially, it establishes a pattern of death which is directly
associated with the three functions where the First receives hanging, the Second
438 25. COMPARATIVE MYTHOLOGY
burning or bloodshed (by sword or other appropriately military weapon), and
the Third Function victim is drowned.
25.10 Death and the Otherworld
There is an abundance of evidence for various beliefs concerning death and the
afterlife in the diVerent Indo-European traditions but ferreting out an original
belief is diYcult. Many Indo-European traditions portray death as a journey
and in the case of Celtic, Germanic, and Greek, and to a lesser extent Slavic and
Indic, this may involve a journey across a river where the deceased is ferried by
a *g
ˆ

erh
a
ont- ‘old man’. On this journey they may also encounter a dog who
serves either as a guardian of the Otherworld or as a guide. Here we have some
linguistic evidence in the cognate names of Greek Ke
´
rberos, the three-headed
dog of Hades, and the Indic S
´
a
´
rvara, one of Yima’s dogs, both deriving from
a PIE *k
ˆ
e
´
rberos ‘spotted’. Both Greek and Indic traditions also have a river
‘washing away’ either memories or sins while Germanic and Celtic traditions
attest a belief of wisdom-imparting waters; Bruce Lincoln has suggested that
these two may be joined together where the memories of the deceased are
washed away into a river but others, lucky enough, may drink of such water
and gain inspiration. The actual afterlife is attested in so many diVerent ways—
as a pleasant meadow, a place of darkness, island, house, walled enclosure—
that it is diYcult to ascribe any particular belief to Proto-Indo-European. The
ruler of the dead, however, may well be the sacriWced twin of the creation myth
as suggested by Indo-Iranian tradition and to a lesser degree by Germanic.
25.11 Final Battle
Celtic, Italic, Germanic, Indo-Iranian, Armenian, and Greek all reveal traces of
an Indo-European eschatological myth, i.e. a myth that describes the end of the
world in terms of a cataclysmic battle, e.g. the Battle of Kurukshetra from the

Maha
¯
bha
¯
rata, the Second Battle of Mag Tured in Irish tradition, Ragnaro
¨
kin
Norse tradition, the Battle of Lake Regillus in Roman history, Hesiod’s Tita-
nomachy, and the Plain of Ervandavan in Armenian history. In all these
traditions the end comes in the form of a major battle in which gods (Norse,
Greek), demi-gods (Irish), or major heroes (Roman, Indo-Aryan, Armenian)
are slain. The story begins when the major foe, usually depicted as coming from
adiVerent (and inimical) paternal line, assumes the position of authority
among the host of gods or heroes, e.g. Norse Loki, Roman Tarquin, Irish
25. COMPARATIVE MYTHOLOGY 439
Bres. In this position he exploits the labour of the protagonists until he is driven
out and returns to his own people. A new leader then springs up among the
protagonists (e.g. Irish Lug, Greek Zeus) often the *nepo
¯
t- ‘grandson’ or
‘nephew’ of the deposed leader. The two sides then prepare for a major war
(in Germanic and Iranian myth there is also a great winter) and the two forces
come together and annihilate each other in a cataclysmic battle. Since a new
order is called into existence after the battle, the myth may not be eschato-
logical in the strict sense but rather represent a mythic encounter that brought
a past golden age to an end.
25.12 Current Trends
Current trends in Indo-European comparative mythology are taking several
directions. The evidence for trifunctional (or quadri-functional) patterns is con-
tinually being augmented by further examples both from well-researched

sources, e.g. Indic, Roman, Norse, and from other traditions such as Greek and
Armenian that have seen far less attention. Moreover, an increasing number of
scholars have been examining the narrative structure of the earliest literary
traditions of the various Indo-European groups to reveal striking parallels
between diVerent traditions. For example, N. B. Allen has shown how much of
the career of the Greek Odysseus is paralleled by distinct incidents in the lives of
Arjuna in the Maha
¯
bha
¯
rata, the Buddha in the earliest Buddhist texts, and
Cu
´
Chulainn in early Irish heroic literature. Other scholars such as Claude
Sterckx, Stepan Ahyan, and Armen Petrosyan have uncovered detailed corres-
pondences inother early Indo-Europeantraditions. According toAllen, the close
coincidences go beyond both the type of random generic parallels that one might
expect between diVerent literary traditions and beyond what we might ascribe to
some formofdistant diVusion. Heargues that suchcomparisonsprovides us with
at least some of the detritus of the Proto-Indo-European narrative tradition.
Further Reading
The best general treatise is Puhvel (1987a); for the core of Dume
´
zil see Dume
´
zil (1968–
73) and Littleton (1973); cases for a ‘Fourth Function’ can be found in Allen (1987), Lyle
(1990); the mythic structure of IE medicine is to be found in Benveniste (1945); the
‘‘three sins of the warrior’’ are the subject of Dume
´

zil (1970); representative new
approaches within the Dume
´
zilian tradition that seek new patterns of underlying
Indo-European narratives include Ahyan (1998), Allen (2000a, 2000b, 2002), Miller
440 25. COMPARATIVE MYTHOLOGY
(2000), Petrosyan (2002), Sterckx (1994); a diVerent approach to IE mythology can be
found in Haudry (1987). The topics of creation, sacriWce, death, and the Otherworld can
be found in the various works of Lincoln (1980, 1981, 1986, 1991, 1995); various deities
are discussed in Dexter (1996), Nagy (1974a), Watkins (1995); the divine twins are
treated in Ward (1968), Lehmann (1988), Grottanelli (1986), Dubuisson (1992), and
York (1995); the subject of sacred vocabulary is handled in York (1993); summaries of
the eschatological model are found in O’Brien (1976) and more recently Bray (2000);
death beliefs are in Puhvel (1969), Hansen (1980), and Lincoln (1980), while burial is
discussed by Jones-Bley (1997).
25. COMPARATIVE MYTHOLOGY 441
26
Origins: The Never-Ending
Stor y
26.1 The Homeland Problem
Sir William Jones had hardly postulated the existence of what we now term the
Indo-European language family before he set future Indo-European studies its
longest and most frustrating problem. In the same lecture (see Section 1.1) in
which he described the relationship between the various ancient languages, he
also remarked that in a future discourse he would attempt to follow them back
to ‘some central country’. In his later lectures he argued that the homeland lay
in greater Iran. This assertion set oV a legacy of debate in which homelands
have been set anywhere from the North to the South Poles, from the Atlantic to
the PaciWc. Before we brieXy review the diVerent approaches and solutions to
the homeland problem, we should ask ourselves whether this is even a legitim-

ate problem.
Why must the Indo-European languages be derived from a smaller geo-
graphical area than that in which we Wnd them when they begin to enter the
historical record? Why couldn’t they have always been there, at least since the
time of Homo sapiens sapiens? This is indeed an argument made by a several
scholars who locate the Indo-Europeans right across Europe from the begin-
ning of the Upper Palaeolithic onwards, i.e. c. 40,000 years ago. The reasons for
not making such an assumption are several.
26.1 The Homeland Problem 442
26.2 Homeland Approaches 444
26.3 What Does the Homeland
Look Like? 453
26.4 Evaluating Homeland
Theories 454
26.5 Processes of Expansion 458
26.6. Where Do They Put it Now? 460
First, from our initial historical records onwards we can see Indo-Europeans
expanding centrifugally, at least beyond the periphery of their historical distri-
bution (Maps 1.1, 1.3). Iberia maintains evidence of both prehistoric and current
non-Indo-European populations, e.g. Basques, as does Italy (Etruscans). The
Iranian language expanded south to absorb the earlier Elamite language of
southern Iran and Indo-Aryan languages spread southwards and eastwards to
absorb, at least partially, Munda and Dravidian languages. The Anatolian
languages are so laced with loanwords from their non-Indo-European neigh-
bours that languages such as Hittite are often seen as having been superimposed
on a Hattic substrate.
Second, the reconstructed lexicon, no matter how narrow or broadly we
interpret it, makes it abundantly clear that the proto-language possessed
a mixed arable agriculture-stockbreeding economy, some metals, ceramic tech-
nology, and wheeled transport. As agriculture did not exist in either Europe or

India prior to the seventh millennium, it is diYcult to sustain an argument that
the Indo-Europeans were scattered across Eurasia from the fortieth millennium
bc onwards. As a cultural phenomenon, Proto-Indo-European cannot have
begun disintegrating until it had already adopted a Neolithic economy and
technology.
Third, the greater an area that we assign to a language (whatever continuum
of dialects that we might imagine for Proto-Indo-European), the greater the
opportunity for language divergence over time. In concrete terms, the larger the
area that we imagine for the speakers of what we notionally reconstruct as
a proto-language, the more rivers, mountains, seas, variation in economic
strategies, social systems, contacts with non-Indo-European substrates, we
must imagine contributing to linguistic diversity. While we cannot assign a
one-to-one relationship between language change, time, and area, we do know
that all of these features are factors. Conversely, if we Wnd a single language
over a large area we tend to presume a short period of time for its spread.
There have been periods of broad consensus, e.g. an Asian homeland was the
favourite for much of the nineteenth century but a European homeland (where
in Europe was another question altogether) has been the primary choice of
most scholars since the early twentieth century. Now, the consensus is still
probably European but there are a number of scholars who would support
Anatolia (Turkey) or other areas of Asia. With so much dispute and with
everyone working with the same general body of evidence, we are clearly
dealing with profound methodological diVerences. How do we determine the
centre of the spread of a language? Are there universal principles that we can
employ to determine the prehistoric location of a language?
The most obvious approach to Wnding the Indo-European homeland, i.e.
selecting a geographical location in time and convincing the rest of the world
26. ORIGINS: THE NEVER-ENDING STORY 443
that one is right, is examining the distribution of languages from their centres in
many historically controlled situations so that we can observe the processes and

principles involved. The problem with this approach is that there is really
nothing suitable. Where we can observe the expansion of a major language
group, e.g. Romance or Germanic, it is under historical circumstances that
are hardly likely to have obtained at the time of Indo-European expansions.
Where we Wnd language families that more closely approximate the social
conditions of Proto-Indo-European, e.g. Chinese, Uralic, Algonquian, we
Wnd ourselves dealing with other unresolved homeland problems. In short, no
language family has provided a suitable laboratory to work out conWdently the
rules of the game. That is not to say that many solutions do not try to argue
from what are posited to be well-established principles, but few if any of such
principles can be regarded as wholly compelling from an empirical standpoint.
26.2 Homeland Approaches
The search for the Indo-European homeland is an exercise in logic and the
diversity of solutions is primarily due to the variety of approaches that have
been taken. Below follows a brief compendium of the type of more serious
arguments that have been adduced to locate the original location of the Indo-
Europeans.
26.2.1 External Language Relations
Just as adjacent languages may mutually inXuence each other when in contact
so also do adjacent language families. Linguists have discerned loanwords or
grammatical loans (or mutual inheritances) between Indo-European on the one
hand and Uralic, Afro-Asiatic (here Semitic), and Kartvelian. These presumed
contacts have supported homelands set in the steppelands of Eurasia (with the
Uralics in the forest zone to the north), in eastern Anatolia (to accommodate an
interface between Kartvelian and Semitic), and in central Asia (distant Semitic
relations and again with Uralics to the north). The problems with such an
approach have been discerning the time depth of the ‘contacts’, i.e. what have
been interpreted as Uralic-Proto-Indo-European loans by some have been seen
to be much later contacts between Iranians or Indo-Iranians and Uralics. The
nature of the contacts may also be disputed, i.e. where we may Wnd apparent

loanwords between two language families, it is presumptive that these must
have been in direct contact with one another when the language groups could
444 26. ORIGINS: THE NEVER-ENDING STORY
still have been geographically distant and the lexical connections are Wander-
wo
¨
rter, i.e. far-travelled cultural loanwords. Third, it may be disputed whether
the relationship reXects a contact relationship between two diVerent language
families or whether the evidence points to the retention of shared terms from
genetically related language families which share a common origin, i.e. the
similarities go back to a time long prior to the formation of the two proto-
languages involved. It should also be emphasized that language families are not
synchronic, i.e. there is no reason to postulate the same time depths to every
language family. Some uniform proto-language may have been spoken over
a geographically compact area at the same time when their neighbours had
already diVerentiated into diVerent language groups of an already expanded
family.
26.2.2 Centre of Gravity
The distribution of the diVerent language groups, it is argued, should provide
important clues as to their origin. In the biological sciences, for example, a map
of the diVerent genera and species of a plant or animal often indicates the
probable area of origin. This argument generally involves an appeal to max-
imum diversity to indicate the centre of a language dispersal. The English
language is most uniform in areas where it has expanded most recently (Aus-
tralia, New Zealand) and shows more evidence of regional dialects in areas
settled somewhat earlier (North America) and greatest diversity in areas where
it has existed longest (England). If we continue this approach, we would argue
that as there are far more Germanic languages in north-west Europe it is far
more likely that English derived from there rather than the reverse, i.e. that the
other Germanic languages spread from England to the Continent. This ap-

proach has been a staple of homeland solutions everywhere in the world. It also
has a converse principle: where we Wnd the greatest homogeneity of languages,
that area is likely to have been most recently occupied. In general, these
principles have selected for homelands in or adjacent to the Balkans. Here we
can list a series of language groups, e.g. Greek, Albanian, Illyrian, Thracian,
Dacian, Slavic, which are portrayed as a central core while on the periphery we
Wnd large areas occupied by single language groups (Indo-Iranians in the east
and Celtic (here seen in terms of its broad Iron Age distribution) in western and
central Europe).
The problem with this approach is that it is extremely diYcult to apply at a
consistent date or with a suitable control of the actual diversity of the languages
involved. We may be able to pack our putative Balkan core with Illyrian,
Thracian, and Dacian but we have no idea how diVerent they were from each
26. ORIGINS: THE NEVER-ENDING STORY 445
other or from neighbouring Indo-European groups. Moreover, we have no
absolute measure of diVerence in the Wrst place. Although we tend to use
languages as the common unit of measurement, the diversity between lan-
guages of the same family is hardly uniform. For example, the major Scandi-
navian languages of Norwegian, Swedish, and Danish are very broadly
mutually intelligible as also are the Eastern Slavic languages of Russian and
Byelorussian, more distantly Ukrainian. If we simply count the number of early
languages we know and their diversity in speciWc locations, it is probable that
Italy would be judged the winner with its numerous, poorly attested Iron Age
languages that shared the peninsula along with Latin. In Italy the linguistic
diversity attested by our earliest linguistic records has been replaced with
relative linguistic uniformity by the spread of Latin. In Anatolia the linguistic
diversity of our earliest records was replaced by the spread of Greek and then,
later, by the spread of Turkish. How many other areas where our earliest
knowledge is of linguistic uniformity are the products of exactly the same
process?

26.2.3 Cladistic Correlation
The family tree of the Indo-European languages has often been seen as a partial
proxy to the geographical relationships between the diVerent languages. For
example, many if not most linguists would see the separation between Anatolian
and the other Indo-European languages as among the earliest ‘splits’. For this
reason, homeland solutions are devised to accommodate these intrafamily
relationships, generally by having the homeland not too distant from the
historical seats of the Anatolian languages. Following this line of reasoning,
the Proto-Indo-European homeland is placed in Anatolia, requiring all the
other Indo-European languages to separate oV from Anatolia (either to the
east or to the west), or the homeland is placed somewhere not too distant from
Anatolia, e.g. the steppelands, so that the future Anatolians might be accounted
for by the initial Indo-European expansions. The problems involved with this
method are several. First, there are competing family trees to explain the Indo-
European languages and the diVerences will govern the nature of the geograph-
ical relationships proposed. Second, it is presumptuous to read geographical
co-ordinates into a linguistic relationship. For example, although many trees
will suggest reasons for placing the Indo-Iranians linguistically close to the
Greeks and Armenians (see Figs. 5.1, 5.2, and 5.3), how do we translate this
relationship into a geographical expression of where they may have shared this
mutual development (or contact)? It may have been in India, Iran, the steppes,
Anatolia, the Balkans, Greece itself, or somewhere outside this broad band.
446 26. ORIGINS: THE NEVER-ENDING STORY
26.2.4 Onomastics
This approach presumes that the proto-language itself might have left identiW-
able traces on the named landscape. The primary use of such an approach has
been in the area of river names on the assumption that these represent the oldest
and lexically least altered component of the landscape. Hence, if one can discern
Proto-Indo-European names for rivers, we can presume that we have found an
area in which the Proto-Indo-Europeans once lived. Such an approach has

thrown up homelands in the Baltic or central and eastern Europe. These hydro-
nymic solutions run into very serious problems. Many would dispute the inter-
pretation of the empirical evidence, i.e. that one can conWdently etymologize the
names of rivers beyond an existing language system. The systems of ancient river
names require appeals not to speciWc Indo-European languages but to deriv-
ations from Proto-Indo-European roots, and there is no way of checking the
credibility of assigning river names like ‘the bright’, ‘the runner’, etc. One
linguist’s Indo-European names become another’s proto-Basque, or Caucasian
or anything else.
There are several other onomastic approaches although these play little part
in more recent research. Iranian tradition spoke of an Airyana vaeja ‘seed of the
Aryans’ as a particular (but unspeciWed) geographical location and that trad-
ition set many scholars oV to localize it in some particular place. Moreover, it
was often assumed by such scholars that the homeland of the ‘Aryans’ could be
assumed, without much further ado, to be the homeland of their ancestors, the
Proto-Indo-Europeans, as well. In actuality the Airyana vaeja would have been
the homeland of (a major branch of) the Iranians alone.
26.2.5 Conservation Principle
One of the recurrent arguments employed to determine the Indo-European
homeland on the basis of purely linguistic evidence is the assumption that
the homeland is most likely in the area where we Wnd the least altered Indo-
European language. This presumption is based on the logic that, if a language
has not moved, it will have experienced far less impetus to change, e.g. impact
of substrates or contacts with other languages, than those languages that have
spread through more distant migration. This principle was initially applied in the
nineteenth century when it was assumed that Sanskrit was the closest to the
proto-language, but over the course of the next century two other contenders
appeared. The archaic nature claimed for Anatolian made it possible to suggest
that it was the least moved language, but this conclusion was mitigated by the
26. ORIGINS: THE NEVER-ENDING STORY 447

clear evidence of loanwords from a variety of its neighbours (Semitic, Hattic,
Hurrian) and the internal evidence that indicated that Hittite had been adopted
by a non-Indo-European substrate. Alternatively, the Baltic languages, particu-
larly Lithuanian, were seen to be remarkably conservative, especially in light of
their late attestation. This conservatism provided one of the cornerstones for
those who sought an Indo-European homeland on the shores of the Baltic Sea.
The conservation principle suVers from several serious defects. Its applica-
tion requires one to measure how conservative were the various Indo-European
languages, but this comparison cannot be done on a level playing Weld because
the various languages entered the historical record at diVerent times. To
compare Sanskrit with a putative date of c.1200 bc with Lithuanian at ad
1800 is patently unfair (and assessing the state of Lithuanian at 1200 bc
requires a time machine). As it is impossible to compare any more than three
language groups at c. 1000 bc (Indo-Iranian, Greek, and late Anatolian) one is
not comparing the full range of Indo-European languages. If one applies the
principle by a time when all the languages can be brought into play, we then
Wnd ourselves comparing the modern languages of India (Hindi-Urdu, Bengali,
etc.) with the Romance languages (French, Italian, etc.) and we will have to
ignore all earlier evidence, including whole language groups (Anatolian, Toch-
arian) or well-attested earlier stages of the language groups (Sanskrit, Latin).
Secondly, there is no empirical measuring device to ascertain in any reliable
quantitative manner how conservative or how innovative the Indo-European
languages are. There is no commonly agreed scale by which one could compare
each language group against a standard (reconstructed Proto-Indo-European).
Third, the underlying logic of the exercise is largely based on the assumption
that language change is a product of language contact, i.e. the reason that
a language spread through migration is likely to experience more change is
that it has undergone imperfect learning by substrate populations (or come into
contact with foreign languages). While these may inXuence language change,
they are hardly the only reasons for it. Finally, if conservation did indicate lack

of movement from a putative homeland we would expect that there would be
a corresponding gradient of conservatism running from the homeland to the
most travelled language group; in fact, there is no such evidence of a graduated
abandonment of the ‘mother tongue’ over distance.
26.2.6 Linguistic Palaeontology
The analysis of the reconstructed proto-lexicon for clues as to the location of
a proto-language is a widely employed technique although many prefer a diVer-
ent term, e.g. lexico-cultural analysis, from the original nineteenth-century term,
448 26. ORIGINS: THE NEVER-ENDING STORY
linguistic palaeontology, that led to discredited results. The underlying premiss
is that if we can reconstruct the environment and technology known to the
Proto-Indo-Europeans, we should be able to determine their location. The
main areas of interest are the words for trees, animals, and material culture, all
of which mayhavehad restricted distributionsinthe past. The techniquerequires
an appeal to archaeological and palaeo-environmental evidence to set broad
limits on where the proto-language may have been spoken. This exercise is often
only intelligible when we also have some idea of when Proto-Indo-European was
spoken (see Chapter 6) because the distribution of plants, animals, and most
especially material culture has varied greatly through time. If one accepts the
broad dates provided earlier, i.e. c. 4500-2500 bc, for Proto-Indo-European, the
lexico-cultural evidence does little to conWne the potential area of the homeland.
The diYculty is that the more geographically speciWc the reconstructed item, the
less likely it is for the word to have survived once the Indo-Europeans expanded
beyond a region where it existed. Or, the word might then be applied to a new
species of plant or animal and we will be left with critical uncertainty as to what
the proto-lexeme actually meant. We have already seen this in three of the classic
Indo-European homeland arguments which required us to determine whether
*lo
´
k

ˆ
s meant ‘Atlantic salmon’ or ‘salmon trout’, *bheh
a
g
ˆ
o
´
s meant the common
beech (Fagus silvatica) or some other species of beech (Fagus taurica or Fagus
orientalis) or some other tree altogether, and whether *h
1
e
´
k
ˆ
wos referred to the
‘domestic horse’ or the ‘wild horse’ (or both)? There is no cultural item that
clinches a homelandinany speciWc location butit should not beimagined that the
lexical cultural evidence is altogether useless. It does provide us with a fairly
consistent impression of the time of Proto-Indo-European (Late Neolithic/
Eneolithic) and it provides us with evidence that renders some potential home-
lands much less likely than others, e.g. the absence of the evidence of the horse
altogether from both Greece and Italy before the Bronze Age makes it less likely
that these were the earliest seats of the Indo-Europeans.
26.2.7 Physical Anthropology
The use of physical anthropological evidence (now the term ‘bio-archaeo-
logical’ is often preferred) emerged as a major technique of the latter nineteenth
century but after the excesses of twentieth-century racists it has few supporters,
at least within the sphere of Indo-European studies, as this area is precisely
where the excesses were inXicted. The assumption here is that human physical

type may serve as proxy evidence for the speakers of a language family. There
were several approaches. One depended on phenotypic diVerences, i.e. the
outward appearance of diVerent peoples. Scholars mined historical records
26. ORIGINS: THE NEVER-ENDING STORY 449
and literature for descriptions of the earliest Indo-Europeans and then argued
whether they were blond or brunette (given the range of meanings of colour
terms in ancient literatures this is not always an easy task) and employed such
evidence to determine the likely homeland. This method produced arguments
of truly staggering illogic as pseudo-scientists sought the epicentre of European
blondness under the assumption that only there could one have acquired light
hair and only there could have been the homeland. As cloning techniques were
unlikely to have been present during the period 4500–2500 bc,itisdiY cult to
see why the phenotype of the original population of so physically disparate
speakers as the Indo-Europeans had to be uniformly blond, brunette, or
whatever colour one might imagine.
A second approach involved the analysis of skeletal anatomy, primarily the
human skull, which was divided into certain ‘subracial’ categories, e.g. Nordic,
Armenian, Mediterranean, or into the broader categories of skull length to
breadth ratio, i.e. brachycephalics (brachycranials if it was your skull and not
your living head) who had wide heads and dolichocephalics (dolichocranials)
with long heads. The problem here is that if children of dolichocephalics could
turn out brachycephalic, how could one seriously regard such broad distinc-
tions as meaningful? It has proven diYcult to sort out which measurements of
the human skull are measuring something that is entirely genetic, i.e. inherited,
versus those which may diVer either randomly or because of the environment,
especially the diet. Those who still measure skulls generally do so within the
context of multivariate analysis where a number of diVerent, and presumably
more reliable, measurements are analysed statistically in order to determine the
direction of gene Xow from one population to another. Even this technique is
not widely employed simply because many, perhaps most, physical anthropo-

logists have abandoned such analysis.
A third approach is genetic, i.e. either the analysis of the genetic composition
of modern populations or the extraction of genetic data (ancient DNA) from
skeletal material. This method has proved to be a growth industry in language
studies (there is grant money out there to be gained) but the results are still far
from reliable. Analysis of modern populations as proxy evidence for past mi-
grations, especially migrations that should have occurred thousands of years
earlier, have yielded quite conXicting interpretations. One of the earliest and
still discussed is the work of Luca Cavalli-Sforza and his colleagues on the
distribution of human genes in European populations where the Wrst principal
component, indicated by a genetic path from South-West Asia westwards across
Europe, has been interpreted as the result of the expansion of the Wrst farmers in
the seventh millennium bc or, alternatively and in no way in association with the
spread of Indo-European speech, that of modern Homo sapiens sapiens popula-
tions c. 40,000 bc. The temptation to read every cline on a map ofgeneticfeatures
450 26. ORIGINS: THE NEVER-ENDING STORY
as a migration and tie it to a putative linguistic movement has led to ostensibly
circular reasoning. As for the use of ancient DNA, actually establishing gene
Xow among ancient populations where there is control for the date of the gene
Xow, the techniques involved are of a far higher magnitude of diYculty. Ancient
DNA is often very poorly preserved, expensive to recover, and without analysis
of a large area, valid conclusions cannot be made. The technique may in time
become a useful tool but that day is some way oV.
Finally, the problem with both genetic and phenetic approaches is that there is
an assumed correlation between language and human physical type. Studies of
current language boundaries do reveal some correlations but many of these
involve natural barriers (seas, mountains) and none can be reliably factored
for time, i.e. there is no way to distinguish whether a currently observable border
between, say, Romanian (Italic) and Bulgarian (Slavic) is a modern feature or
reXective of an earlier border between Dacian and Thracian or a still earlier

border. The requirement of a genetic trail could only be accepted if one required
that for language shift to occur there must be a constant human vector involved
so that there was major directional gene Xow. Given the fact that inmostcaseswe
are probably speaking of language shift between neighbouring peoples, there is
no requirement whatsoever that the trail of language shift should also leave
a clearly deWned genetic trail as well. Nor for that matter can we assume that if we
do Wnd a genetic trail, this necessarily resulted in a language shift favourable for
those carrying the gene rather than their absorption by local populations.
26.2.8 Retrospective Archaeology
We have already seen archaeological involvement in the use of linguistic
palaeontology but it may be employed in a number of other ways as well. The
most obvious is the retrospective method where one examines those archaeo-
logical cultures that must have been associated with diVerent Indo-European
language groups and attempts to work backwards to the ‘proto-culture’. The
unit of analysis here is the so-called ‘archaeological culture’, a classiWcation
device employed by archaeologists to deal with similar and geographically
conWned material culture and behaviour. This method fails to convince for at
least two major reasons.
The retrospective technique presumes that one can employ cladistic tech-
niques to provide an archaeological family tree much like a linguistic tree. But
this is not at all what one actually does because the archaeologically deWned
cultures show constant mutual contact in terms of ornamental styles, architec-
ture, metallurgy, or any other phenomenon of cultural life, i.e. there is no single
line of ‘gene Xow’ within a continuum of archaeological cultures. Moreover, the
26. ORIGINS: THE NEVER-ENDING STORY 451
deWnition of the individual units may well vary through time, e.g. in the
Neolithic ceramics tend to be critical for distinguishing one culture from
another but by the Bronze Age, metallurgical tradition and mortuary practice
become more critical elements.
Secondly, even if one were convinced of the underlying logic of the retro-

spective method, it still falls apart on empirical grounds once one has worked
back to c.3000 bc (in some cases the retrospective method disappears al-
together). Many of the language groups of Europe, i.e. Celtic, Germanic,
Baltic, and Slavic, may possibly be traced back to the Corded Ware horizon
of northern, central, and eastern Europe that Xourished c.3200–2300 bc. Some
would say that the Iron Age cultures of Italy might also be derived from this
cultural tradition. For this reason the Corded Ware culture is frequently
discussed as a prime candidate for early Indo-European; in the past it was
even suggested as the Proto-Indo-European culture. However, the Corded
Ware cannot even remotely explain the Indo-European groups of the Balkans,
Greece, Anatolia, nor those of Asia. For the steppeland regions of Eurasia, the
retrospective method takes us back through the Bronze Age Andronovo and
Timber-grave cultures of the Eurasian steppe to the underlying Yamna culture
of c.3600–2200 bc. This method can supply us with an archaeological proxy for
the Eastern Iranians but that is about all the retrospective method gets us. We
may argue that the Yamna culture should minimally reXect the proto-Indo-
Iranians if not more; however, we cannot do this by the retrospective method
since there is no ancestral culture that territorially underlies the Iranians or
Indo-Aryans, i.e. there is no speciWc culture X that both embraces the historical
seats of the Indo-Iranians and can also be traced back to the Yamna culture.
Similarly, there is really no solid evidence in the retrospective method in Greece
that takes us anywhere that we can conWdently tie to one of the other two
‘ancestral cultures’; nor Anatolia. Sooner or later the retrospective method
leads us to a series of what seem to appear to be independent cultural phenom-
ena that somehow must be associated with one another. In that lies most of the
archaeological debate concerning Indo-European origins.
26.2.9 Prospective Archaeology
The opposite method to a retrospective approach is a prospective approach
where one starts with a given archaeological phenomenon and tracks its
expansion. This approach is largely driven by a theory connected with the

mechanism by which the Indo-European languages must have expanded.
Here the trajectory need not be the type of family tree that an archaeologist
might draw up but rather some other major social phenomenon that can move
452 26. ORIGINS: THE NEVER-ENDING STORY
between cultures. For example, in both the nineteenth century and then again
in the later twentieth century, it was proposed that Indo-European expansions
were associated with the spread of agriculture. The underlying assumption here
is that only the expansion of a new more productive economy and attendant
population expansion can explain the widespread expansion of a language
family the size of the Indo-European. This theory is most closely associated
with a model that derives the Indo-Europeans from Anatolia about the seventh
millennium bc from whence they spread into south-eastern Europe and then
across Europe in a Neolithic ‘wave of advance’. A later alternative mechanism
is the spread of more pastoral societies who exploited the horse (and later the
chariot) and carried a new language across Europe and Asia from the fourth
millennium bc onwards. The underlying assumption here is that the vector of
Indo-European language spread depended on a new, more aggressive social
organization coupled with a more mobile economy and superior transportation
technology. As this theory sets the homeland in the steppelands north of the
Black and Caspian seas among diVerent cultures that employed barrows for
their burials (Russian kurgan), it is generally termed the Kurgan theory.
Although the diVerence between the Wave of Advance and Kurgan theories
is quite marked, they both share the same explanation for the expansion of the
Indo-Iranians in Asia (and there are no fundamental diVerences in either of
their diYculties in explaining the Tocharians), i.e. the expansion of mobile
pastoralists eastwards and then southwards into Iran and India. Moreover,
there is recognition by supporters of the Neolithic theory that the ‘wave of
advance’ did not reach the peripheries of Europe (central and western Medi-
terranean, Atlantic and northern Europe) but that these regions adopted
agriculture from their neighbours rather than being replaced by them.

In short, there is no easy way to locating the Indo-European homeland; there
is no certain solution.
26.3 What Does the Homeland Look Like?
One of the problems of homeland research is that often those searching for it are
not clear what they are looking for or likely to Wnd. If we consider the problem
from Wrst principles, then there is absolutely no reason to imagine that Proto-
Indo-European began with the origins of human speech. Once that is accepted,
then obviously Proto-Indo-European must have had ancestral stages that
pre-date its appearance. In some cases, linguists have attempted to reconstruct
Pre-Proto-Indo-European, generally through internal reconstruction. Often the
ancestry is traced to earlier proposed linguistic stages, e.g. Proto-Indo-Uralic or
Nostratic, but even here one is seldom proposing a language stage earlier than
26. ORIGINS: THE NEVER-ENDING STORY 453
c.15,000–10,000 bc. Moreover, as we trace Indo-European along the develop-
mental line of a still longer language tree, our control of time and space becomes
increasingly weaker. If one, for example, wished to derive Proto-Indo-European
from Nostratic, there is an overwhelming temptation to locate a Nostratic
homeland and use this as a proxy homeland for Proto-Indo-European. But
once this is done, we exclude from the equation vast tracts of Eurasia whose
cultures will then remain linguistically anonymous for they fall outside the
geographical area of anyone’s Nostratic (generally localized to somewhere in
South-West Asia). We are accumulating unknowables at an alarming rate.
The result is that Proto-Indo-European deWnes that stage in a linguistic
continuum retrievable by the comparative method. It was not an ‘instant’ in
the life of a language nor was it a recognizable event to those who spoke it
(occasionally in the nineteenth century scholars provided explicit scenarios
where the Proto-Indo-Europeans resided in some conWned, possibly isolated,
territory where they ‘perfected’ their language). If we must accept that the
temporal boundaries of our deWnition are blurred over many centuries, perhaps
on the order of one or two thousand years, then it follows that the territorial

boundaries of the proto-language are also very blurred. It is almost inconceiv-
able that the linguistic borders of Proto-Indo-European could have remained
static for a millennium or two. The best we can hope for is a dead reckoning of
an area at a particular range of time in the hope that it encompasses much of
what we believe to have been the ancestral speech of the Indo-Europeans.
26.4 Evaluating Homeland Theories
In a world with so many competing theories, how can we evaluate which are the
most probable? Many homeland solutions depend on the reiteration (often in
tones of vastly greater conWdence than is warranted) of one or two pieces of
evidence and selective amnesia concerning all the objections to the theory.
Although there is not a single solution that may not be regarded as damaged
goods, there are some that seem beyond repair, but we need some explicit
guidelines to separate these from the real contenders. The following comprises
a partial arsenal of criteria by which one might assess a potential solution.
26.4.1 Temporal Relationship
A solution cannot date after 2000 bc by which time we may expect to Wnd an
already diVerentiated Anatolian as well as Indo-Iranian and probably Greek.
454 26. ORIGINS: THE NEVER-ENDING STORY
How early a solution is admitted depends on individual decisions regarding the
temporally most diagnostic vocabulary. That the vocabulary is clearly one
reXecting at least a Neolithic economy and technology, i.e. domesticated plants
and animals, ceramics, means that it cannot be set anywhere on this planet
prior to c. 8000 bc. Although there are still those who propose solutions dating
back to the Palaeolithic, these cannot be reconciled with the cultural vocabu-
lary of the Indo-European languages. The later vocabulary of Proto-Indo-
European hinges on such items as wheeled vehicles, the plough, wool, which
are attested in Proto-Indo-European, including Anatolian. It is unlikely then
that words for these items entered the Proto-Indo-European lexicon prior to
about 4000 bc. This is not necessarily a date for the expansion of Indo-
European since the area of Proto-Indo-European speech could have already

been in motion by then and new items with their words might still have passed
through the continuum undetected, i.e. treated as inheritances rather than
borrowings. All that can be concluded is that if one wishes to propose
a homeland earlier than about 4000 bc, the harder it is to explain these items
of vocabulary.
26.4.2. Linguistic Relationship
Any solution should accommodate the broad requirements of whatever family
tree is being proposed. In general, there is probably some broad although not
universal consensus that would see a separation between Anatolian and the
other Indo-European languages (see Figs. 5.3 and 5.4). Many have argued that
Greek, Armenian, and Indo-Iranian share a number of innovations that sug-
gest that there should have been some form of linguistic continuum between
their predecessors. This line of thinking then presupposes various peripheries
such as Germanic, Baltic, and Slavic in some form of relationship and possibly
Celtic and Italic in another, still related to the north European languages. The
position of Tocharian still remains beyond solid consensus other than the fact
that it cannot be brought into the same continuum as Indo-Iranian. If a
solution to the homeland can avoid totally contradicting these relationships,
it can be regarded as a potential model.
26.4.3 External Relationship
There is evidence for loanwords and possibly genetic connections between
Proto-Indo-European and other language families, most particularly Uralic
26. ORIGINS: THE NEVER-ENDING STORY 455
and Semitic. The interpretation of the empirical evidence here is not now (nor
ever has been) the subject of much consensus and attempts to dead reckon the
Proto-Indo-European homeland on a notional idea of its relationship with
these other language families have plenty of problems. At best a solution
should be able to devise a way by which Proto-Indo-European could have
borrowed from and loaned words to these two major groups. It would, how-
ever, be a mistake to imagine that these relations can be translated into speciWc

geographic co-ordinates, especially when we do not know the prehistoric
location of the other language families any better than Indo-European.
26.4.4 Total Distribution Principle
The correct solution to the Indo-European homeland problem explains the
origins and distribution of all the Indo-European languages. All too often a
solution proceeds from some form of argument for the local continuity of
a language in a particular area and then extrapolates this back to the homeland
itself. In the nineteenth and Wrst half of the twentieth century, the model of
continuity helped drive a north European solution to the homeland problem,
i.e. if there is no evidence that anyone brought a new language into northern
Europe, then there must have been local continuity in this region and all the
other Indo-European languages derive from northern Europe. Today there is
an entire school that makes a similar argument for local continuity in northern
India and argues that there lies the homeland. In both cases—or any other case
for regional continuity—a solution is made for one area and the rest of
the Indo-European world is forced to accommodate it, generally without the
slightest credible evidence. No solution is valid if it only rests on local continu-
ity; it must provide a viable model for the spread of all the Indo-European
languages.
26.4.5 Plausible Vector Principle
The expansion of the Indo-European languages was a social phenomenon or
many individual phenomena that spanned much of Eurasia. This expansion
could not have taken place without a social vector that should have left some
trace in the archaeological record (ancient DNA may eventually have some
role to play here). Generally, all solutions can be divided into two main
models: demographic replacement and language shift. In the Wrst, the primary
vector will be a new population speaking some form of Indo-European that
456 26. ORIGINS: THE NEVER-ENDING STORY